Journal of Catalysis 268 (2009) 175–179
Contents lists available at ScienceDirect
Journal of Catalysis journal homepage: www.elsevier.com/locate/jcat
Research Note
Identification of the OH groups responsible for kinetic basicity on MgO surfaces by 1H MAS NMR Céline Chizallet a,b,1, Hugo Petitjean a,b, Guylène Costentin a,b, Hélène Lauron-Pernot a,b,*, Jocelyne Maquet c,d, Christian Bonhomme c,d, Michel Che a,b,e a
UPMC Univ Paris 06, UMR 7609, Laboratoire Réactivité de Surface, F-75005 Paris, France CNRS, UMR 7609, Laboratoire Réactivité de Surface, F-75005 Paris, France UPMC Univ Paris 06, UMR 7574, Laboratoire de Chimie de la Matière Condensée, F-75005 Paris, France d CNRS, UMR 7574, Laboratoire de Chimie de la Matière Condensée, F-75005 Paris, France e Institut Universitaire de France, France b c
a r t i c l e
i n f o
Article history: Received 11 May 2009 Revised 27 August 2009 Accepted 6 September 2009 Available online 8 October 2009 Keywords: 1 H MAS NMR MgO Surface hydroxyl groups Basicity Methylbutynol Catalysis Spectroscopy
a b s t r a c t Depending on the localization of hydroxyl groups on MgO surface irregularities, their topology and coordination can differ. In order to discriminate their reactivity, three MgO samples with different morphologies are hydroxylated at various levels. From 1H MAS NMR three kinds of OH groups can be distinguished. It is shown that the catalytic activity of base sites in the conversion of 2-methylbut-3yn-2-ol is governed by species giving a 1H NMR signal at d < 0.7 ppm, assigned to low coordinated O1CH and O2CH formed by water dissociation on steps and corners. Their reactivity is discussed considering charge and orbital analysis from DFT. Ó 2009 Elsevier Inc. All rights reserved.
1. Introduction Examples showing the role of hydroxyl groups (OH) in heterogenous catalytic reactions proceeding on base sites can be found in the literature and are of growing interest because of applications in fine chemistry and biomass valorization. In some cases, improvement of the catalytic results is obtained upon addition of water to the reactant flow like that in aldolic condensation [1], isopropanol decomposition, [2] or oxidative coupling of methane [3]. Oxides such as MgO can also be modified by appropriate pretreatment such as thermal treatment [4], hydration, [5,6] or hydrogenation [7] to exhibit surface OH groups active in isomerization [4], H–D exchange, [6] and alcohol conversion [5,7]. A recent review by Corma and Iborra [8] gives an excellent account of the results obtained on alkaline earth oxides and hydroxides. Moreover, basic OH groups of hydrotalcites have been shown to be catalytically active in aldolic condensation [9,10], Michaël [11] and Claisen–Schmidt [12] condensations, and transesterifications [13]. * Corresponding author. Fax: +33 01 44 27 60 33. E-mail address:
[email protected] (H. Lauron-Pernot). 1 Present address: IFP-Lyon, Direction Catalyse et Séparation, BP3, 69390 Solaize France. 0021-9517/$ - see front matter Ó 2009 Elsevier Inc. All rights reserved. doi:10.1016/j.jcat.2009.09.003
We investigated earlier the relationship between the thermodynamic Brønsted basicity of a surface, i.e., its ability to displace the deprotonation equilibrium of an acidic molecule and the reactivity of the base sites [5,14]. It was found that, despite a lower Brønsted basicity, hydroxylated surfaces are more reactive in 2-methylbut3-yn-2-ol (MBOH) conversion, used as model reaction [15], than clean ones. To first order, two main kinds of hydroxyl groups are expected 2 from hydroxylation of Mg2þ LC OLC pairs: one resulting from protonation of surface oxide ions O2 LC and one resulting from hydroxylation of the magnesium cation Mg2þ LC . They can interact via Hbonding but can also be isolated. Moreover, depending on their location on steps, corners, kinks or divacancies, the OLCH groups may exhibit different coordination numbers (L = 1, 2, 3, or 4) and thermal stabilities [16]. The aim of this paper is to identify which kind of OLCH group is acting as a base site in MBOH conversion in order to help understanding why, despite a weak Brønsted basicity, they exhibit a high reactivity. In earlier works, it was shown that FTIR could not afford the identification of active sites [17] probably because different kinds of OLCH groups vibrate in the narrow band obtained on hydroxylated MgO surfaces [18] and that their contribution cannot be accurately extracted by spectra decomposition. 1H MAS NMR is
176
C. Chizallet et al. / Journal of Catalysis 268 (2009) 175–179
also an interesting method to characterize OH groups but, in the case of zeolites, the data are in good correlation with IR spectroscopy [19]. Nevertheless some discrepancies to this correlation have already been pointed out in the literature especially in the case of Ca2+ and Mg2+-exchanged zeolites [19]. Moreover, as theoretical calculations showed, in the case of MgO, that some species not distinguishable by IR spectroscopy can be differentiated by 1H MAS NMR [20], it was necessary to check whether 1H NMR could afford a distinct evaluation of active sites. Thus, different MgO samples partially hydroxylated were characterized by 1H MAS NMR and by catalysis of the MBOH conversion. In order to rationalize the reactivity of the different OH groups, theoretical calculations of the charge distributions and HOMO/LUMO orbitals energy are performed from a natural population analysis.
tained at 303 K. It has been checked by varying the mass of the sample and the inert flow rate that no diffusional limits could be observed under these conditions. Because acetone and acetylene are the only products detected, catalytic data are expressed in terms of conversion only. Embedded cluster geometry optimizations were performed at the DFT level, within the B3LYP [23,24] hybrid exchange correlation functional, using the Gaussian03 code [25]. Hydroxylated clusters modeling the various surface OLCH groups were previously investigated to obtain their vibrational [18], nuclear [19], and electronic [26,27] properties. The 6-311+G** basis set was used except for third order neighbors (Mg) of hydroxyl groups for which Mg LANL2 effective core potential was used to avoid unphysical polarization with the embedding species (from 913 to 2100 point charges, see Ref. [18]). Charge distributions and HOMO/LUMO orbital occupations were evaluated by a natural population analysis.
2. Experimental 3. Results and discussion In order to perform a quantitative analysis of NMR spectra, in a preliminary work (not reported here), a Hahn-echo procedure was compared to the single pulse procedure. It was concluded that the Hahn-echo procedure helps to localize the probe and the empty rotor signals on the single pulse spectrum but that, to be accurate, it implies a very important number of scans and prohibiting time. Thus, for quantitative analysis, the single pulse procedure was selected, and the broad bands relative to protons of the probe and rotor were modeled by Gaussians on the spectrum acquired with the empty rotor in the same conditions. 2D NOESY NMR experiments [20] showed that at least five maxima can be observed on the spectra thus helping their decomposition as shown in Fig. 1. Assignment of these signals to OLCH groups formed upon water adsorption on MgO defects (see Table 1) was made on the basis of theoretical calculations [20]. It has been shown that three chemical shift domains can be distinguished depending on the nature of the OLCH groups – dH > 0.7 ppm, attributed to most of H-bond donor groups corresponding to signals at 0.8, 0.0, and 0.3 ppm, Fig. 1. – dH 0.7 ppm for the O3CH and O4CH isolated groups (on kinks and divacancies) and for some H-bond donor groups (O4CH on monatomic steps). This contribution is picked up at 0.8 ppm, Fig. 1.
6.0
4.0
2.0
0.0
-0.9 -1.4
0.0 -0.3
-0.8
Experimental signal Decomposition Sum of the contributions
0.8
As described earlier [21], MgO samples were prepared by thermal decomposition at 1273 K in vacuum of Mg(OH)2 precursors leading to three materials such as MgO-precipitation, MgO-hydration, and MgO-sol–gel. The BET specific surface areas were found to be 167 m2 g1 for MgO-precipitation and 150 m2 g1 for both MgO-hydration and MgO-sol–gel. It was checked that there was no surface reconstruction upon hydroxylation treatment [5,21]. The hydroxyl coverage of the samples was controlled by a pretreatment step with identical conditions before 1H MAS NMR and catalysis measurements. The samples are ‘‘cleaned up” at 1023 K under nitrogen flow (20 cm3 min1). Then, they are cooled down to 373 K and contacted with water vapor flow (P H2 O ¼ 840 Pa, flow rate = 20 cm3 min1) for 10 min. They are then dehydroxylated at intermediate temperature (473 K < T < 1073 K) in a nitrogen flow (20 cm3 min1) for 2 h. In the following, only the final dehydroxylation temperature N that determines the degree of hydroxylation of a sample will be given: MgO-sol–gel-673 is a sol–gel sample that was dehydroxylated at N = 673K. For NMR experiments, the pretreatment of 125 mg of sample was performed in a ‘‘U shape” quartz cell equipped with two valves allowing complete isolation of the system (sample + cell). The pretreated MgO sample system was transferred into a glove box (argon atmosphere; H2O < 1.5 ppm; O2 < 3 ppm) where 4 mm (external diameter) zirconia rotors were filled and closed with Kel-F caps. The spectra were recorded at room temperature with a Bruker Avance 400 spectrometer at 9.4 T, equipped with a 4 mm probe and with a spinning rate of 12.5 kHz. A 90° single pulse of 3.3 ls was used with a recycling delay of 5T1 over a spectral window of 250 kHz. T1 was estimated for each sample and each pretreatment temperature, using a Saturation Recovery sequence. The receiver gain remained constant at a value of 200 throughout the experiments, and 128 free induction decays per spectrum were accumulated. The chemical shifts were determined by reference to an external tetramethylsilane (TMS) sample. The signal of the empty rotor was recorded at the same conditions for each experiment. For all spectra, the intensities were normalized to the same specific surface area and the decomposition of the signal was performed with the DMFIT program [22]. The MBOH catalysis reaction was performed as described elsewhere [5]. Wafers were prepared from catalyst powders and crushed into pellets of 125 to 200 lm in diameter. Isosurface area experiments were performed adapting the mass of catalyst loaded in the reactor so as to obtain a surface of 5 m2 and diluting with SiC (Prolabo, 250 lm diameter) so as to finally reach 75 mg of solid. It was checked that SiC alone does not convert MBOH. After pretreatment, the reactor was cooled to the reaction temperature of 393 K. The MBOH partial pressure (PMBOH = 3.3 kPa) was adjusted by bubbling gaseous nitrogen (100 cm3 min1) in liquid MBOH main-
-2.0
-4.0
-6.0
δ (ppm) Fig. 1. 1H MAS NMR spectrum (single pulse sequence), corresponding decomposition, and fit of MgO-sol–gel sample after treatment in flowing nitrogen (20 cm3 min1) at 1023 K for 1 h, hydration at 373 K (about 840 Pa water diluted in nitrogen (20 cm3 min1) for 10 min), and subsequent treatment in flowing nitrogen (20 cm3 min1) at 673 K for 2 h.
177
C. Chizallet et al. / Journal of Catalysis 268 (2009) 175–179 Table 1 Location, schematic representation, and spectroscopic feature of the different kinds of OH groups modeled earlier [16,18,19]. Location
a
Molecular description
Monoatomic step
b
Location
H H O4c
O1c
O4c
Mg4c H
Valley H
O1c
O4c
Mg3c H
H
O2c
O3c
Mg4c
Mg2þ 3C -terminated kinks H
Mg5c
O3c
O4c
Mg4c Mg3c Mg5c
Divacancy
H
Mg5c H
O2c Mg5c
H
O2 3C -terminated kinks
H
b
H
corners
Mg5c H
O2 3C -terminated corners
Molecular description
O2c Mg4c
Edge
O5c
a
Mg2þ 3C -terminated
H
H
O1c
H
O3c
Mg4c
O3c
O3c Mg4c Mg3c Mg5c
a
Nomenclature used in former publications. When H-bonds form, donor OH groups are represented in red and acceptor OH groups in blue. (For interpretation of the references to colour in this table, the reader is referred to the web version of this article.) b
– dH < 0.7 ppm for O1CH and O2CH groups corresponding to signals at 0.9 and 1.4 ppm, Fig. 1. The large signals shown in Fig. 1 are due to probe and rotor contributions. The same decomposition was made on the different samples obtained by dehydroxylation at temperatures varying from 673 to 973 K. The conversion of MBOH was then compared to the integrated area of the NMR contributions at dH > 0.7 ppm, dH 0.7 ppm, and dH < 0.7 ppm. No relationship was found between the MBOH conversion and the amount of OH groups giving a contribution to the two first domains (d > 0.7 ppm and d around 0.7 ppm). On the reverse, a good correlation (Fig. 2) is clearly established between the amount of protons giving a signal at d < 0.7 ppm and MBOH conversion. From the assignment of this signal to O1CH or O2CH groups a schematic representation of these different kinds of OH groups is given in Scheme 1. The OLCH (L = 3 or 4) counter part can be found in the two domains corresponding to dH P 0.7 ppm explaining why there is MBOH Conversion (%) 70 60
MgO-hydration MgO-sol-gel
50
MgO-precipitation
40 30 20 10 0 0
200
400 600 800 1000 1200 1400 Area (a.u.) of NMR deconvoluted lines normalized per specific area of the catalysts at δ < -0.7 ppm
Fig. 2. MBOH conversion in isosurface experiments as a function of the integrated area of the NMR lines corresponding to domain: d < 0.7 ppm. The experimental points are given for the three MgO samples at different levels of hydroxylation (N varying from 673 to 973 K).
H
H
H
O1c
H
OLC
MgL’C
OLC
H bond acceptor - O 1CH
H H
O2c MgL’C
MgL’C
H bond acceptor - O 2CH
OLC
O2c MgL’C
MgL’C
Isolated - O2CH
Scheme 1. Molecular description of different catalytically active sites.
no correlation between the catalytic activity and the area of the NMR line located at either dH > 0.7 ppm or dH = 0.7 ppm independently (data not shown). The identification of the active sites by 1H NMR allows to understand why these peculiar OH groups cannot be clearly distinguished with FTIR [17] The narrow IR band at 3740 cm1 contains the contributions of H-bond acceptors (O1CH and O2CH) identified as active sites by NMR, but also all the isolated OLCH. Thus, its area cannot be representative of the amount of active sites. Now that the active sites have been identified, the mechanism of the reaction on these OH groups can be considered. This is of great interest when considering that if water has often a positive effect in base catalysis by direct addition to the reactants or indirectly by working with a hydroxylated sample, its role played in the mechanism is seldom discussed. In order to explain the reactivity observed on MgO in a H–D exchange reaction, Hoq et al. [6] proposed a concerted mechanism involving both surface O2 and OH groups: the surface O2 traps the proton (or deuterium) of the alkane molecule while a surface OH group releases a proton. However, in such a cooperative effect between a basic O2 and an acidic OH group (Scheme 2a), a maximum of activity should be expected for partially hydroxylated surfaces, whereas a continuous decrease of MBOH conversion is observed with increasing dehydroxylation temperature and thus for decreasing the OH coverage [5]. So a two-step mechanism in which the first deprotonation step is not necessarily rate-determining (Scheme 2b) can be considered in which the OLCH groups act as Brønsted base sites. In the case of zeolites, the basicity has been often linked to the electrostatic charge on oxygen [28]. Calculation of charge distributions and energies of the HOMO 2p orbitals of O atom are given in Table 2.
178
C. Chizallet et al. / Journal of Catalysis 268 (2009) 175–179
H (a)
+
+ OH
H
O
O
Mg
O2-
Mg
H
+
O2-
O +
+ H
H
H OH OH + (b)
Mg
H 2+ OH +
O k
Mg
H
+
Mg
O
+
+
HO H
H
H O +
H
(c)
O
H
Mg
O
-
OH2
H
Mg
2+ O2-
O +
+ H +
OH
H
Mg
O
H
-
Scheme 2. Different kinds of mechanisms for the conversion of MBOH on OH catalytically active sites.
Table 2 Charge distribution on O and H atoms and energy (Haa) of the 1s orbitals of H atoms and 2p orbitals of O atoms for OH groups on hydroxylated surfaces evaluated by DFT. System
OL0 C —H
Type
qO
E2p(O)
qH
E1s(H)
Type
qO
E2p(O)
qH
E1s(H)
O2C–H O1C–H O2C–H O1C–H
1.368 1.352 1.345 1.373
0.323 0.253 0.274 0.207
0.470 0.472 0.473 0.457
0.148 0.221 0.204 0.250
O4C–H O4C–H O5C–H O3C–H
1.432 1.470 1.524 1.425
0.371 0.330 0.356 0.258
0.511 0.520 0.516 0.513
0.126 0.169 0.098 0.238
O1C–H
1.384
0.263
0.473
0.211
O4C–H
1.447
0.349
0.522
0.152
Mg2þ 3C -terminated kinks
O3C–H O3C–H
1.386 1.389
0.352 0.394
0.481 0.481
0.142 0.095
O3C–H O4C–H
1.434 1.399
0.382 0.413
0.489 0.485
0.114 0.057
O2 3C -terminated kinks
O2C–H
1.359
0.086
0.450
0.337
O3C–H
1.384
0.173
0.480
0.321
Monoatomic step Edge Valley O2 3C -terminated corners Mg2þ 3C -terminated corners Divacancy
a
Mg–OLC–H
1 Ha = 27.211 3845(23) eV.
The absolute electrostatic charge on oxygen for a given system is more negative for OH groups formed by protonation of surface oxygen (OLCH) than for those formed by hydroxylation of Mg2+ ions (Mg–OLCH with L = 1, 2 or 3) (all negative values). According to this criterion, the former should be stronger base sites whereas they are not the active sites. On the other hand, the energy of the 2p orbital of O atom of OLCH groups with L = 1 or 2 which have been identified as active sites is higher than the others. These groups can thus be considered as more reactive toward protonation, and so more nucleophilic. The key parameter governing the reactivity toward MBOH conversion could thus be the nucleophilic character of OH groups. Another hypothesis has to be considered because as the lesscoordinated active OLCH groups (L = 1 or 2) all have a vicinal OH group (Scheme 1), even when they are not H-bonded and called isolated (in the case of O2 3C -terminated kinks). Thus another plausible explanation for the high reactivity of hydroxylated surfaces is the existence of a concerted mechanism (Scheme 2c) involving two vicinal OH groups: one H-donor (O1CH or O2CH) and one Hacceptor (O3CH or O4CH) even if not involved in H-bonding. This hypothesis is also in agreement with the nature of the active sites identified on the surface because the inactive sites are the OH
groups located in divacancies and Mg2+-terminated kinks that exhibit only OLCH vicinal groups with L P 3.
4. Conclusion The combined use of NMR and catalysis as well as theoretical calculations has led to a refined description of OH groups of MgO catalytically active in the conversion of MBOH. A correlation has been found between the catalytic activity and the amount of OH groups giving a NMR signal at d < 0.7 ppm. Thus the catalytically active sites could be identified as O1CH and O2CH formed by hydroxylation of the steps, corners, and O2 3C -terminated kinks. From calculations of the energy of the 2p orbital of O atoms of OH groups, it appears that the active sites have the highest energy, in line with the higher reactivity of these sites and related with what is usually called ‘‘nucleophily”. Moreover, all the identified active sites have a vicinal OH counterpart that could easily behave as an acid. Thus a concerted mechanism can be considered to explain the high reactivity of these OH groups despite their poor base strength. Work is in progress to evaluate the most plausible mechanism from a theoretical standpoint.
C. Chizallet et al. / Journal of Catalysis 268 (2009) 175–179
Acknowledgment The authors acknowledge the financial support of ANR Project: BASICAT, ANR-05-JCJC-0256-01.
References [1] G. Zhang, H. Hattori, K. Tanabe, Appl. Catal. 36 (1988) 189. [2] J.A. Wang, X. Bokhimi, O. Novaro, T. Lopez, R. Gomez, J. Mol. Catal. A 145 (1999) 291. [3] S. Kus, M. Otremba, A. Torz, M. Taniewski, Fuel 81 (2002) 1755. [4] J.L. Lemberton, G. Perot, M. Guisnet, J. Catal. 89 (1984) 69. [5] M.L. Bailly, C. Chizallet, G. Costentin, J.M. Krafft, H. Lauron-Pernot, M. Che, J. Catal. 235 (2005) 413. [6] M.F. Hoq, I. Nieves, K.J. Klabunde, J. Catal. 123 (1990) 349. [7] T. Yamakawa, M. Takizawa, T. Ohnishi, H. Koyama, S. Shinoda, Catal. Commun. 2 (2001) 191. [8] A. Corma, S. Iborra, Adv. Catal. 49 (2006) 239. [9] K.K. Rao, M. Gravelle, J.S. Valente, F. Figueras, J. Catal. 173 (1998) 115. [10] F. Prinetto, D. Tichit, R. Teissier, B. Coq, Catal. Today 55 (2000) 103. [11] B.M. Choudary, M. Lakshmi Kantam, C.R. Venkat Reddy, K. Koteswara Rao, F. Figueras, J. Mol. Catal. A 146 (1999) 279. [12] M.J. Climent, A. Corma, S. Iborra, A. Velty, J. Mol. Catal. A 182–183 (2002) 327. [13] A. Corma, S.B.A. Hamid, S. Iborra, A. Velty, J. Catal. 234 (2005) 340. [14] C. Chizallet, M.L. Bailly, G. Costentin, H. Lauron-Pernot, J.M. Krafft, M. Che, Catal. Today 116 (2006) 196. [15] H. Lauron-Pernot, Catal. Rev. – Sci. Eng. 48 (2006) 315–361. [16] C. Chizallet, G. Costentin, M. Che, F. Delbecq, P. Sautet, J. Phys. Chem. B 110 (2006) 15878.
179
[17] C. Chizallet, G. Costentin, H. Lauron-Pernot, J.M. Krafft, P. Bazin, J. Saussey, F. Delbecq, P. Sautet, M. Che, Oil Gas Sci. Technol. 61 (2006) 479. [18] C. Chizallet, G. Costentin, M. Che, F. Delbecq, P. Sautet, J. Am. Chem. Soc. 129 (2007) 6442. [19] E. Brunner, H. Karge, Z. Phys. Chem. 176 (1992) 173. [20] C. Chizallet, G. Costentin, H. Lauron-Pernot, M. Che, C. Bonhomme, J. Maquet, F. Delbecq, P. Sautet, J. Phys. Chem. C 111 (2007) 18279. [21] M.L. Bailly, G. Costentin, H. Lauron-Pernot, J.M. Krafft, M. Che, J. Phys. Chem. B 109 (2005) 2404. [22] D. Massiot, F. Fayon, M. Capron, I. King, S. Le Calve, B. Alonso, J.O. Durand, B. Bujoli, Z. Gan, G. Hoatson, Mag. Res. Chem. 40 (2002) 70. [23] A.D. Becke, J. Chem. Phys. 98 (1993) 5648. [24] C. Lee, W. Yang, R.G. Parr, Phys. Rev. B 37 (1988) 785. [25] M.J. Frisch, G.W. Trucks, H.B. Schlegel, G.E. Scuseria, M.A. Robb, J.R. Cheeseman Jr., J.A. Montgomery, T. Vreven, K.N. Kudin, J.C. Burant, J.M. Millam, S.S. Iyengar, J. Tomasi, V. Barone, B. Mennucci, M. Cossi, G. Scalmani, N. Rega, G.A. Petersson, H. Nakatsuji, M. Hada, M. Ehara, K. Toyota, R. Fukuda, J. Hasegawa, M. Ishida, T. Nakajima, Y. Honda, O. Kitao, H. Nakai, M. Klene, X. Li, J.E. Knox, H.P. Hratchian, J.B. Cross, V. Bakken, C. Adamo, J. Jaramillo, R. Gomperts, R.E. Stratmann, O. Yazyev, A.J. Austin, R. Cammi, C. Pomelli, J.W. Ochterski, P.Y. Ayala, K. Morokuma, G.A. Voth, P. Salvador, J.J. Dannenberg, V.G. Zakrzewski, S. Dapprich, A.D. Daniels, M.C. Strain, O. Farkas, D.K. Malick, A.D. Rabuck, K. Raghavachari, J.B. Foresman, J.V. Ortiz, Q. Cui, A.G. Baboul, S. Clifford, J. Cioslowski, B.B. Stefanov, G. Liu, A. Liashenko, P. Piskorz, I. Komaromi, R.L. Martin, D.J. Fox, T. Keith, M.A. Al-Laham, C.Y. Peng, A. Nanayakkara, M. Challacombe, P.M.W. Gill, B. Johnson, W. Chen, M..W. Wong, C. Gonzalez, J.A. Pople, Gaussian 03. Revision C.02, Gaussian Inc., Wallingford CT, 2004. [26] C. Chizallet, G. Costentin, H. Lauron-Pernot, J.-M. Krafft, M. Che, F. Delbecq, P. Sautet, J. Phys. Chem. C 112 (2008) 16629. [27] C. Chizallet, G. Costentin, H. Lauron-Pernot, J.-M. Krafft, M. Che, F. Delbecq, P. Sautet, J. Phys. Chem. C 112 (2008) 19710. [28] D. Barthomeuf, G. Coudurier, J.C. Védrine, Mat. Chem. Phys. 18 (1988) 553.